Hyperhomocysteinemia and cognitive impairment both predict mortality and partly because of dietary associations. We have hypothesized that for, nutritional reasons, homocysteine and cognition may act jointly to determine elder survival. In a Nutrition and Health Survey in Taiwan (1999-2000), some 1412 representative elderly were followed up for mortality up to 10 years. Cognition was assessed by the Short Portable Mental Status Questionnaire. Food and B vitamin intakes with their biomarkers, and plasma homocysteine, were measured at baseline. The possible effects of cognition on homocysteine-associated mortality were ascertained with Cox proportional-hazards models. Homocysteine was higher in those who were older, male, and single, consumed less fish and tea, and with alcohol and smoking. In models adjusted for these variables, when homocysteine exceeded 14.5 μmol/L, mortality was 1.80-fold more than when <9.3 μmol/L (hazard ratio [HR], 1.80; 95% confidence interval [95% CI], 1.20-2.71). P for trend was 0.002 and interactive with sex (P < .002). However, these homocysteine-mortality associations were dependent on cognition (P = .03); adjustment for food intake or nutrient status made little difference. Homocysteine did not predict cognitive impairment (adjusted OR, 1.40; 95% CI = 0.50-3.93). Vitamins B(1), B(2), and B(6) accounted somewhat for cognitive impairment. Cognition predicted mortality, fully adjusted for available covariates and also for homocysteine (HR, 3.66; 95% CI, 1.64-8.20) but interactively with homocysteine. Thus, the B-group vitamin insufficiency and cognitive impairment associations with premature mortality are confirmed. Yet cognition is inter-related with homocysteine in its association with survival in ways not detectably altered by foods or food-derived vitamins.